about the receptive fields calculated in your paper

peiyunh / tiny

Tiny Face Detector, CVPR 2017

https://cs.cmu.edu/~peiyunh/tiny

Other

1.13k stars 320 forks source link

about the receptive fields calculated in your paper #69

Closed caney1989 closed 6 years ago

caney1989 commented 6 years ago

hello: I have read your paper carefully and I'm confused about the receptive fields corresponding to res-50 calculated in your paper. I have calculated it by myself, the results are 35, 99, 291, corresponding to res2, res3, res4, different from the results in your paper. Could you tell me the details about this calculation process?? Thank you !!

peiyunh commented 6 years ago

Matconvnet has a implementation for computing the receptive field sizes. You can refer to matconvnet/matlab/+dagnn/@DagNN/getVarReceptiveFields.m